ABSTRACT: This dissertation examines several grammatical features of Potawatomi, a Central Algonquian language, whose syntactic distributions in traditional narrative are different from those found in everyday discourse. These grammatical features include the verbal paradigmatic orders known as the independent and conjunct, a verbal prefix é-, and obviation. In everyday discourse, independents are main clause forms, and conjuncts are generally subordinate clause forms. The verbal prefix é- is a marker of factivity within a subordinate clause. In narrative, however, most main clause verbs take the conjunct prefixed with é-. The function of obviation in everyday discourse is largely syntactic, with several several obligatory contexts of application. In narrative, however, it is optionally used to foreground and background characters, and to represent shifts in viewpoint.
These distributions raise the issue of the relationship between between syntactic structure and discourse structure, and present the challenge to linguistic theory of accounting for syntax that is dependent on discourse context. I argue that the discourse-dependent distributions of these grammatical phenomena can be explained in a cognitive linguistic framework, which assumes that syntax is not autonomous, but part of a continuum of form / meaning pairings which includes the lexicon and discourse structures. Within this framework, I propose that the aspects of Potawatomi grammar described above participate in several constructions that map a particular grammatical form onto multiple functions in both syntax and discourse. Using Mental Spaces theory, I show that these functions are related to each other in the way they structure and index aspects of mental spaces networks.
I also argue for a productive mental space blend in Potawatomi that takes as its input spaces syntactic and discourse uses of constructions. In this way, possible contexts for the application of a construction in one domain can be associated with established contexts in the other. When the cross-space mappings are made, the blend can be 'run' and the construction applied to the new domain. This blend demonstrates that a full semantic description of these constructions requires explaining their functions within the domains of syntax and discourse, as well the relationships between their functions across these domains.